Driverless Networks

Project Description

Automated network management has been the holy grail of network management research for decades with the aim of closing the management loop and achieving autonomous networking, i.e., networks capable to autonomously monitor their status, analyze potential problems, make control decisions, and execute corrective actions. There have been several attempts to achieve self-managing networks, including policy-based management, autonomic networking, knowledge-driven networks, and recently self-driving networks. However, practical deployments have largely remained unrealized. Several limiting factors can be attributed to this, including the existence of many stakeholders with conflicting goals, reliance on proprietary hardware and a complex web of interacting protocols, lack of global visibility restricting network-wide optimizations, and the inability to process network telemetry at scale.

The stars are now aligned to realize the vision of autonomous networking thanks to advances in network softwarization, recent breakthroughs in machine learning, and the availability of cloud platforms for large-scale data processing. However, these individual technologies are merely pieces of a bigger puzzle yet to be solved for the successful realization of autonomous networks. A number of challenging issues need to be addressed not only to create the synergy between these different technology domains but also to develop a fundamentally new approach for the orchestration and management of softwarized networks. These include the ability to program the data plane in a protocol-independent manner for adaptive monitoring and control policy enforcement, real-time processing of streaming monitoring data, predictive machine learning for closed-loop network management, orchestration algorithms for cost-effective, resilient, and efficient service provisioning.

Sponsors and Partners

NSERC